Image Resolution Upscaling via Two-Layered Discrete-Time Cellular Neural Network

This paper proposes a novel image resolution up-scaling method using discrete-time cellular neural network (DT-CNN) with multi-level quantization function for output of a cell. The nonlinear interpolative approximation capability of the DT-CNN is used to generate an resolution enhanced image from its low-resolution version. Our proposed method consists of two-layered DT-CNN. At the first layer stage, the DT-CNN is used to obtain the optimal weight parameter which makes possible to represent the original image with a weighted linear combination of finite impulse response function such as Gaussian function and wavelet function. At the second layer stage, the image obtained after transition of the first layer DT-CNN is upsampled with arbitrary size, then the resolution enhanced image is obtained by the convolution with the B-template which is derived by extending the A-template spatially. The experimental evaluation shows that the proposed method produces better results than the conventional image resolution enhancement methods